This manuscript is based on dissertation research completed under the direction of Dr. Jie-Qi Chen at Erikson Institute, and was funded by a Head Start Research Scholars Grant from the U.S. Department of Health and Human Services, Administration for Children and Families and a grant from the Chicago Public Schools. Running head: PCK FOR PRESCHOOL MATHEMATICS PCK for Preschool Mathematics: Teacher Knowledge and Math-Related Language Contribute to Children’s Learning Jennifer S. McCray Erikson Institute
54
Embed
Running head: PCK FOR PRESCHOOL MATHEMATICSmathcoaching.erikson.edu/.../04/McCray-PCKforPreschoolMathematics.pdfErikson Institute. PCK for Preschool ... PCK for Preschool Mathematics:
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
This manuscript is based on dissertation research completed under the direction of Dr. Jie-Qi Chen at Erikson Institute, and was funded by a Head Start Research Scholars Grant from the U.S. Department of Health and Human Services, Administration for Children and Families and a grant from the Chicago Public Schools.
Running head: PCK FOR PRESCHOOL MATHEMATICS
PCK for Preschool Mathematics: Teacher Knowledge and Math-Related Language
Contribute to Children’s Learning
Jennifer S. McCray
Erikson Institute
PCK for Preschool 2
Abstract
This study examines relationships between a new teacher interview designed to assess
teachers’ pedagogical content knowledge (PCK) for preschool mathematics, their math-
related language, and school-year gains in mathematics ability scores by children in their
classrooms. Twenty-six teachers and 141 children in Head Start programs in a large
Midwestern city participated. Analysis using hierarchical linear modeling (HLM) finds
significant relationships between scores on the new PCK Interview and gains in
children’s mathematics learning. In a new finding, frequency of teacher math-related
language is only significantly related to child outcomes when delivered outside the large
group setting; frequency of math-related language during “circle time” showed no
association. A final HLM model combining PCK Interview scores and teacher math-
related language outside the large group finds both contribute significantly and positively
to child gains, suggesting that teacher PCK is not entirely mediated through math-related
language as measured here.
PCK for Preschool 3
PCK for Preschool Mathematics: Teacher Knowledge and Math-Related Language
Contribute to Children’s Learning
In 1988, Deborah Ball’s dissertation demonstrated the importance of assessing
teachers’ pedagogical content knowledge (PCK) for teaching mathematics. Ball took
Shulman’s (1986) idea – that a particular kind of content knowledge is specially suited to
teaching – and used it as the basis for a mathematics interview. In it, teachers were
presented with elementary classroom scenarios in which a student misunderstood
something, and teachers were asked how they would respond. Teachers with good PCK
for teaching mathematics, Ball reasoned, would come up with representations of the
mathematics that clarify meaning and guide future thinking. Shockingly few of the
teachers Ball interviewed had even adequate responses to these scenarios; some teacher
responses were simply incorrect. In a further application of Ball’s interview, Ma (1999)
interviewed Chinese teachers. Though these teachers were less educated than their U.S.
counterparts, they offered rich, connected responses, demonstrating what Ma termed a
“profound understanding of fundamental mathematics” (p. xxiv). Ball’s interview
accomplished two important things for mathematics education: first, it provided clear
evidence that U.S. teachers’ PCK for elementary mathematics teaching was lacking; and
second, it suggested the types of content and pedagogy that might constitute good PCK,
providing clear direction for elementary teacher education.
Unfortunately, Ball’s interview and subsequent work related to PCK for
elementary mathematics offer little to those of us wishing to better understand preschool
mathematics and its effective teaching. There are profound differences in the knowledge
bases of elementary and preschool mathematics; in particular, elementary-level
PCK for Preschool 4
mathematics uses written notation to support mathematical thinking, enabling the
accomplishment of longer strings of procedures and acting as a sort of meta-cognitive
crutch. Preschool children are active mathematics learners, but not ready to understand,
let alone use, written arithmetic. Instead, three- and four-year-olds are primed to notice
and explore the quantitative relationships in the world around them, and to begin using
language and other forms of primary representation to crystallize information about
pattern, shape, space, size, and number. Because of this, the preschool pedagogical task
shifts from ensuring a connected, conceptual understanding of new mathematical
procedures such as “moving the decimal over one place,” to helping young children
recognize, name, and begin to experiment with the mathematics in their classroom
environment and their own actions upon it. These particularly-preschool instructional
tasks require a knowledge base distinct from the one Ball’s work has helped to identify.
Meanwhile, converging evidence from neuroscience, developmental psychology,
mathematics and science educators, and early childhood professionals has helped to
create an emerging sense that preschool mathematics education in the U.S. is in need of
substantial improvement. It has become clear that early education is a more important
teaching opportunity than was previously understood. Recent research clearly
demonstrates the profound impact of early environmental influences on long-term brain
development (Markezich, 1996), suggesting the potential of early education to create
long-term benefits. Additionally, policymakers have become aware that more U.S.
children are attending early care and education programs before formal schooling than
ever before (U.S. Department of Education, 2000) and that good early education can have
evidence suggests that higher quality teaching of mathematics in the earliest years of a
child’s education offers the best hope for improving mathematics achievement in the U.S.
(Bowman et al., 2001; Clements, Sarama, & DiBiase, 2004).
While the education literature indicates that teacher quality is the most effective
predictor of student achievement (Darling-Hammond, 2000; Just for the Kids and The
Southeast Center for Teaching Quality, 2002; National Commission on Teaching &
America’s Future, 1996; Rice, 2003), the quality of teaching at the preschool level is
PCK for Preschool 6
extremely variable (Copple, 2004), and mathematical teaching is no exception. A joint
position statement on preschool math by the National Association for the Education of
Young Children and the National Council of Teachers of Mathematics (NAEYC, 2005) is
specific about the problem. The two organizations note the general lack of good teacher
preparation in mathematics and point out that this under-preparation contributes to poor
math-related attitudes among many early childhood teachers who lack confidence in
mathematics anyway. While lack of confidence does not, in itself, prevent a teacher from
teaching math, it can feed an unfortunate tendency to avoid math in the classroom. A
recent contribution to preschool mathematics teacher training goes so far as to devote a
section to this issue, noting “Math Anxiety – You Can Handle It” (Smith, 2001, p.2), and
when surveyed (Carpenter, Fennema, Peterson, & Carey, 1988), both pre- and in-service
teachers in early childhood classrooms expressed great reluctance to teach mathematics,
making comments like “I don’t do math.” Describing early childhood educators, Copley
notes “to them, mathematics is a difficult subject to teach and one area that they often
ignore except for counting and simple arithmetic” (2004, p. 402). There is good reason to
assume that the teaching of preschool mathematics could be improved.
PCK for Preschool Mathematics – Construction of the Teacher Interview
In an attempt to better illuminate what preschool mathematics is and make clear
the importance of teaching it knowledgeably and well, the author designed a new teacher
interview to assess PCK for preschool mathematics. Literature on elementary
mathematics education, preschool pedagogy, and mathematics content for very young
children were consulted to create a tool that might allow teachers to demonstrate the
knowledge that undergirds excellent preschool mathematics teaching. Where there was a
PCK for Preschool 7
lack of relevant preschool educational literature, ideas from cognitive developmental
theory and recent research on quantitative development in early childhood were
incorporated. The concept of pedagogical content knowledge (PCK), however, provided
the guiding framework for the interview’s content and construction.
According to Lee Shulman, who first identified pedagogical content knowledge
(PCK) during his 1985 presidential address to the American Educational Research
Association (Shulman, 1986), PCK is “a knowledge of subject matter for teaching which
consists of an understanding of how to represent specific subject matter topics and issues
appropriate to the diverse abilities and interest of learners” (Shulman & Grosman, 1988,
p. 9). For example, PCK emphasizes knowledge of which content ideas are more central
to the subject and how they connect to one another (subject matter understanding),
appropriate examples for illustrating those concepts (teaching techniques), and awareness
of how those concepts develop in the thinking of novices with differing levels of
experience (knowledge of student development). In other words, knowledge of content,
teaching practice, and student development are effectively combined where good PCK
exists. See Figure 1 for a Venn diagram illustrating how these three bodies of knowledge
overlap to construct PCK.
The new Preschool Mathematics PCK interview assessed here (see Appendix A)
borrows Deborah Ball’s 1988 innovation of situating interview questions within teaching
scenarios and asking teachers to report what specific actions they would take in each
situation (see also Study of Instructional Improvement, 2002). Using teaching scenarios
rather than items solely about content knowledge contextualizes questions so they are
more like the kinds of dilemmas teachers actually encounter in the classroom. This
PCK for Preschool 8
approach provides a mechanism for tapping content knowledge, pedagogy, and
understanding of student learners simultaneously, acting as a better approximation of
PCK than a purely content-related question could ever supply.
Figure 1. Pedagogical Content Knowledge (PCK)
Subject Matter Understanding
Knowledge of Students’
Development
Teaching Techniques
Pedagogical Content
Knowledge
Importantly, Ball’s questions are also designed to assess teachers’ ability to help
students of elementary mathematics maintain and enhance connections between
procedures and concepts in their mathematical thinking. The idea that there are two
distinct types of mathematical knowledge – procedural knowledge, which encompasses
the forms, rules, and processes that make it possible to accomplish mathematical tasks,
and conceptual knowledge, which embodies richly connected ideas about the
mathematical relationships between things and actions – is an old one. U.S. mathematics
education has a long history of emphasizing first one of these types of knowledge, and
then the other, with little effective result (see, e.g., McLellan & Dewey, 1895; Thorndike,
1922; Wheeler, 1939). It was not until the early 1980s that Resnick and Ford suggested
PCK for Preschool 9
that rather than trying to determine their relative importance to math education, the
relationships between concepts and procedures ought to be emphasized (1981). In
alignment with this thinking, Ball’s interview pointedly asks elementary mathematics
teachers to come up with real-world representations that can illustrate the usefulness and
meaning of specific mathematics procedures.
For example, in one interview scenario, pre-service teachers were presented with
the division statement 1 ¾ / ½ (“one and three quarters divided by one half”). Ball
asked the pre-service teachers to “develop a representation – a story, a model, a picture, a
real-world situation” of this division statement (Ball, 1988, p.16). An appropriate model
of this statement would demonstrate that it will answer the question “how many halves
are there in one and three quarters?” The effective teacher, Ball reasoned, would be able
to describe a situation in which such a question is meaningful. As an example of a useful
representation, Ball suggests “A recipe calls for ½ cup of butter. How many batches can
one make if one has 1 ¾ cups butter? Answer: 3 ½ batches” (1988, p. 16). Interview
questions such as this made clear which teachers had more of the kind of knowledge that
would enable them to keep mathematical procedures meaningfully connected to the ideas
they are meant to represent.
The concept-procedure literature in mathematics education is ambivalent about its
own usefulness for preschool mathematics. Interestingly, it defines preschool
“procedures” and elementary (and above) “procedures” extremely differently, though the
ramifications of these differences are largely unexamined (see, e.g., Hiebert & Lefevre,
1986). In the elementary-based literature procedures are often algorithmic processes
enacted upon symbols such as “borrowing from the 10s column,” while in the preschool
PCK for Preschool 10
version of this analysis procedures are generally physical actions taken and enacted upon
concrete objects. For example, in the concept-procedure literature, a toddler uses the
“procedure” of placing a spoon in each teacup to construct a working concept of one-to-
one correspondence (see Sinclair & Sinclair, 1986). From this perspective, very young
children’s mathematics concepts are completely dependent upon their procedures;
mathematical knowledge is being created through action and has not yet been separated
from the world of things. Sinclair and Sinclair comment “the young child cannot do
without actual experience when logico-mathematical knowledge is in its beginnings”
(1986, p. 63). As in elementary mathematics, connections between concepts and
procedures in mathematical thinking are central; however, instead of being at risk of
becoming separated from one another, perhaps they are at risk of not being firmly
established in the first place.
The relatively sparse current literature on early mathematics education seems to
support this perspective. Clements notes “although young children possess rich
experiential knowledge, they do not have equal opportunities to bring this knowledge to
an explicit level of awareness” and suggests the importance of differentiating between
“the intuitive, implicit, conceptual foundation for later mathematics” and the subsequent
elaboration that produces something more like conceptual math knowledge (2004, p. 11).
Copley seems to agree, suggesting that conceptual learning requires something beyond
actions taken upon concrete objects alone. She notes “Early childhood educators say that
children learn by doing. The statement is true, but it represents only part of the picture.
In reality a child learns by doing, talking, reflecting, discussing, observing, investigating,
listening, and reasoning” (2000, p. 29).
PCK for Preschool 11
If preschoolers can benefit from assistance in establishing and consolidating
initial connections between their budding mathematical ideas and more generalized
concepts, this provides an initial portrait of what good preschool mathematics teaching
must be. Following the thinking of children as they interact with materials, recognizing
the mathematical potential in their activities, and knowing how to comment on and
extend their mathematics-related thinking must all be central. To provide opportunities
for teachers to demonstrate these abilities, the new interview presents teachers with two
classroom-based free play scenarios – one in the dramatic play area and one in the block
corner. First, teachers read the scenarios to themselves, then the researcher reads them
aloud, and finally, the researcher asks the teacher to do various things, including:
identify specific math-related topics the children’s play addresses; suggest a comment
they might make to help the children think about/become more aware of the mathematics
in their play; and propose a question they could ask that might encourage children to
experiment with the mathematics in their play and extend their thinking. In this way, the
interview is meant to elucidate preschool teachers’ abilities to help young children
construct and consolidate their initial mathematical concepts.
Each of the mathematics content strands proposed by the National Council of
Teachers of Mathematics (NCTM) in its Principles and Standards for School
Mathematics (2000) is represented in each scenario either through the use of specific
materials, the comments children make during the scenario, or the problems children
encounter and actions they take. For example, Scenario 1 presents teachers with the
following text:
Brittany and Jacob are playing in the dramatic play area and want to put their five babies to bed. There are no doll beds, so they make “cribs” out
PCK for Preschool 12
of three shoeboxes. Jacob says “but there aren’t enough cribs.” Brittany responds, “these babies are younger” picking out the three babies with no hair and setting them near the shoeboxes. She picks up the two babies with thick hair, says “these babies don’t need to nap anymore,” and sets them aside. Jacob says “OK, but this baby needs the most room” and puts the biggest bald baby in the biggest shoebox. Brittany watches him and then puts the medium-sized bald baby in the medium-sized shoebox and the smallest bald baby in the smallest shoebox. Jacob says “now go to sleep, babies.” (Appendix A)
In this scenario, the small, medium, large relationships of the babies and shoeboxes
represent a repeating pattern, while the idea of using a shoebox for a crib is an example of
three dimensional geometric thinking. Recognizing that there “aren’t enough cribs”
requires one-to-one correspondence, which is foundational to number sense. When
Brittany solves this dilemma by putting aside the two babies who have hair, she classifies
and sorts a single set (babies) into two sets (bald babies and babies with hair).
Measurement skills are activated when Jacob determines which baby needs the most
room, and which shoebox offers it. In the interview, teachers who can see more
mathematics in such play and generate effective comments that point it out and
encourage its elaboration score more points.
A small group of experts in quantitative development and preschool pedagogy
assessed the face validity of the interview, and contributed to a list of possible teacher
responses. The list was used to construct response rubrics for each interview question
which allow interviewers to check off key items as they are mentioned by teachers. Early
piloting among three teachers indicated more and different responses than anticipated.
Response rubrics were revised and resubmitted to the experts for comment; their
comments were incorporated. Further piloting among six preschool teachers suggested
PCK for Preschool 13
the interview was sensitive enough to produce sufficient variability with a range of scores
from 12 to 34 and a standard deviation of 7 (scores between 0 and 86 being possible).
Teacher Math-Related Language
The study also explores the relationship between this construction of preschool
mathematics PCK and teachers’ practices by examining teacher math-related language.
Recent work (Ehrlich, 2007; Klibanoff, Levine, Huttenlocher, Vasilyeva, and Hedges,
2006) has found a positive association between frequency of preschool teachers’ use of
mathematics-related talk and gains in mathematics their students make during the school
year. Assuming this finding would be replicated, coding systems from Klibanoff, et al.,
(2006) were applied to the classroom speech of the teachers interviewed so the predictive
power of the interview could be compared to that of math-related language (see
Appendix B).
The conceptualization of PCK above, however, suggested further differentiation
of the relationship of math-related language to child outcomes might be possible.
Teacher math-related language can appear at many times during a preschool day and in
many different settings. It is often a part of the whole classroom meeting, or circle time,
when preschoolers may count the number of children present or review the calendar and
count off the date. It can also, however, be present when teachers work with a small
group of children to complete a project, as in “each ladybug gets two eyes,” or when
teachers are supporting children’s play, as in “four of those cups are little, and two are
big.” Since the representation of good mathematics pedagogy developed above
suggested children develop mathematical concepts while operating directly on the world
around them, it was theorized that circle time might not be the best opportunity for
PCK for Preschool 14
mathematical language to aid in this endeavor. Teacher math-related language offered
while children are actively engaged in their own pursuits, however, could be differently
received and processed because aligned with children’s actions or mathematical
“procedures.” To assess whether pedagogical setting had any bearing on the relationship
between math-related language and child learning, teacher speech data was gathered in
both large group settings (circle time) and in free play and small group settings (non-
circle time).
Research questions were:
1) Is there a significant association between teachers’ PCK interview scores
and gains children in their classrooms make?
2) Is there a significant association between teachers’ math-related speech
and gains children in their classrooms make (replicating the findings of
others)? Is the strength of this association essentially the same for math-
related speech during circle time and outside the circle time setting?
3) Is the relationship between PCK and child gains mediated by teacher
math-related language, and if so, to what extent?
Methodology
Participants
Twenty-six Head Start teachers from a large urban area and 113 children from
their classrooms with assessment scores at two time points participated. To be
considered for the study, teachers had to work in classrooms in which English was not a
second language for a majority of students and be “Head” teachers in their classrooms.
Teachers were selected at random from among this group and recruited for participation;
PCK for Preschool 15
14 CPS teachers agreed to participate. The author randomly selected non-CPS Head Start
agencies to contact from a list provided by City of Chicago staff. Twelve non-CPS Head
Start teachers agreed to participate. Participating teachers were given a $100 stipend and
their classrooms received a set of picture books after teacher interviews were complete.
Classroom size ranged from 13 to 21 children, with a mean of 18 and a standard
deviation of 2.5. While Head Start admits children between the ages of 3-0 and 4-11 at
the start of the school year, on average, classrooms in the sample had five three-year-olds
in September of 2006. Overall, 70% of the students in participating classrooms were
African-American, 23% were Latino, and the remaining students were of other ethnic
origin, including Brazilian, Caucasian, Chinese, Czech, German, Hmong, Polish, and
Vietnamese. Taking overall classroom composition into account, more than half the
classrooms served an all African-American population, several served a mostly-Latino
population, and the remainder served a very diverse population. On average, 23% of
students in these classrooms spoke English as a second language. The 26 classrooms
were clustered within 19 sites. Fourteen sites held only one participating classroom, four
sites had two participating classrooms, and one site held four.
In order to participate, children had to be between 3-4 and 5-0 years old, and be
fluent in English, according to the teacher. When possible, six such children were
randomly selected to participate from among those whose parents had consented in each
classroom. One hundred forty-one children were assessed at T1 in the 26 classrooms that
participated for the length of the study, or an average of 5.4 children per classroom.
These children averaged 4 years and 4 months of age at first testing, and 76 of them, or
53.9%, were female. One hundred ten children, or 78.1%, were African-American, 26
PCK for Preschool 16
were of Latino origin, and 5 were identified as Asian/Pacific Islander. Of these 141
children, only 113 had mathematics assessment scores at two time points (others having
moved away or dropped out of their class). Only these children are included in the
analysis, resulting in an average of 4.3 children per classroom.
The missing subjects were generally demographically similar to the entire sample.
They averaged 4 years and 2 months of age, 90% came from English-only homes, and
their ethnic make-up was strikingly consistent with the sample as a whole, with 80% of
the group African-American, 16% Latino, and 4% Asian. They were, however, more
likely to be female: 73.9% of these T1-only subjects were female, compared with 53.9%
of the sample as a whole. Because of this, the remaining sample includes a higher
percentage of male children (53%) than the original sample (46.1%).
Teacher Language Samples
Participating teachers’ speech was tape-recorded and coded on one randomly
selected day during the period from the beginning of January 2007 until end of April
2007. One hour of teacher speech was recorded and coded, always before noon, and
included both “circle-time” (teacher-led large group) and the period immediately
following; recording/coding continued until 60 minutes were captured. Activity after
circle time consisted of combinations of free play, in which teachers supported children’s
activities in areas of their own choosing, and small-group activities, in which teachers
met with a small group of children to guide them through a teacher-led set of activities.
To facilitate recording, each teacher wore a wireless microphone, which broadcast a
signal to be coded and recorded. Two researchers were present during recording to code
teacher math-related language responses; a timer was utilized to prompt researchers to
PCK for Preschool 17
take turns coding in 10 minute intervals. Teachers were unaware of the mathematics-
focus of the study when the language sample was recorded and coded.
The 60 minutes of coded teacher language began with the teacher’s first address
to the whole group of gathered children at the beginning of circle time. Coding
categories from Klibanoff, et al. (2006) were used to code teacher math-related language
(see Appendix B for category definitions). This coding system heavily emphasizes
references to number and quantity – geometric and spatial math-related language is not
explicitly represented. Because live coding of math-related language using this coding
system had not been previously attempted, all language samples were recoded by the
graduate student researcher using audiotapes, yielding an inter-rater reliability score of
96.4%; disagreements were settled by discussion.
Since language data was collected on only one day for each classroom, an
examination of total math-related language instances was conducted to screen for
outliers. That is, teachers who chose to emphasize mathematics on the one day of
language recording far beyond what they would generally do needed to be accounted for,
since the purpose of the language data was to find a representative sample of math-related
language for each classroom. Using the commonly accepted formula of
< Q1 – (1.5 x IQR) or > Q3 + (1.5 x IQR)
in which Q1 = 1st quartile score, Q3 = 3rd quartile score, and IQR = Inter-quartile range
(3rd quartile score – 1st quartile score), four teachers’ frequency of math-related language
scores were determined to be outliers; these four teachers and the data retrieved from
within their classrooms (e.g., child mathematics gains) were removed from the data set
for all further analysis.
PCK for Preschool 18
Upon review of the audiotapes of these outlier teachers, it was discovered that one
teacher spent the entire free play time going over number flash cards with children,
another played a 25-minute game of number bingo, a third used almost 80% of the 60
minutes to play “find the number” with a small group of children (a game in which she
said, for example, “where is the two?” and children picked out the correct numeral from
a display), while the fourth spent her entire non-circle time working on ordinal
relationships with a small group of children (saying, for example, “three, four, five...what
comes next?”). While these activities may not represent unexpectedly advanced
mathematics teaching, the dramatic impact they have on these teachers’ language
instances on the single day of recording and coding make the scores unlikely to be
representative of their usual practice. Further, since these outlier classrooms represent
the most math-related language delivered outside the circle time setting, their removal
makes the assessment of non-circle time language as a predictor of child outcomes a
more stringent one.
Test of Early Mathematics Ability
To assess gains Head Start children made in mathematics during the study period,
the Test of Early Mathematics Ability, 3rd edition, (TEMA-3) (Ginsburg & Baroody,
2003) was administered twice – once in the fall and once in the spring. The TEMA-3 is a
standardized, norm-referenced instrument designed to assess both informal (non-school
taught) and formal mathematics achievements among 3 to 9 year old children. Basals
and ceilings are utilized in the TEMA-3, and testing begins with an item corresponding to
the child’s age and ends when the child makes five consecutive errors. Children are
asked to count the objects on a card, select the card with more objects and point to it,
PCK for Preschool 19
show a specific number of fingers, perform basic level mental addition and subtraction,
and write numbers, among other things. TEMA-3 has two parallel forms with established
test-retest reliability; children were randomly assigned Form A or Form B at T1 and
received the alternate form at T2.
In 2002, the TEMA-2 (Ginsburg & Baroody, 1990) was reviewed as part of a
workshop on “school readiness” sponsored by the U.S. Department of Health and Human
Services, and was found to be “based on current theory and research…grounded in
normative data…(and) a good predictor of mathematics achievement” (U.S. Department
of Health and Human Services, 2002, p. 23). The normative sample for this assessment
was representative of the national population in terms of sex, race (black, white, and
other), region of the country, residence in an urban or rural community, and parent
occupation (white-collar, blue-collar, service, farm, or other). Cronbach’s Alpha for
TEMA-3 math ability scores at age four is reported to be .93 for Form A and .95 for
Form B, well within the most desirable range for reliability. Test-retest reliability is
similarly high, with a delayed alternate-form coefficient of .93 (Ginsburg & Baroody,
2003).
Testing took place in a quiet setting near the child’s classroom, often at a small
table in the hallway just outside the room. Sometimes a separate room, down the hall
from the classroom, was the option provided by the school. Assessments were conducted
by graduate students in child development. Most TEMAs took approximately 15 minutes
to administer, though assessment times ranged from about 2 minutes to almost 47
minutes, depending upon the child’s ability and willingness to participate.
Results
PCK for Preschool 20
Is There a Significant Association between Teachers’ PCK Interview Scores and Gains
Children in Their Classrooms Make?
Interview reliability. Because the Preschool Mathematics PCK Interview is a new
measure, analyses were conducted to assess its internal reliability before examining its
relationship to teacher practices. As a first measure of reliability, correlations were run
between total Scenario scores. Scores for Scenarios 1 and 2, each meant to reflect PCK
across mathematics content areas, correlated strongly (r = .616, p < .001). Utilizing sub-
scenario scores, Cronbach’s alpha for the items comprising these two Scenarios was a
strong .76, representing an acceptable level of inter-item reliability (according to
Nunnaly, 1978, the standard is .7.) Interviews were coded live, and fifteen (50 %) were
coded simultaneously by two researchers. Inter-rater reliability was 92.8%;
disagreements were settled by discussion and used an interview audiotape. Teachers
scores ranged from 6 to 32 points, with an average score of 21.6 and a standard deviation
of 5.8.
Children’s gains in mathematics. At Time 1, the 91 children in non-outlier
classrooms scored between 65 and 132 on the TEMA-3, with a mean score of 85.11 and a
standard deviation of 12.43. This mean score is nearly 1 standard deviation below the
population average of 100, indicating the sample children were doing noticeably worse
on mathematics than their same-age peers nationally. At Time 2, scores had not changed
that much overall. The children scored between 62 and 124, with a mean of 86.11 and a
standard deviation of 14.74. In fact, six children showed no change in TEMA-3 score
between time points, and 71.4% had scores that changed less than 10 standard points in
either direction. Further, 53.8% of the sample showed either no change in score, or a
PCK for Preschool 21
negative change in score. Recall that the TEMA-3 is standardized for age so that simply
getting the same test items correct from one session to the next is not enough to maintain
the same standardized score. The lack of change reflected in some of these scores may
represent an actual increase in knowledge and/or more items correct on the assessment
while also indicating no movement relative to national averages for same-age peers.
Regardless, this sample of Head Start children is not keeping up nationally, and by this
measure, does not show an overall trend to improve its relative standing.
Relationship between interview scores and child gains. A 3-level Hierarchical
Linear Model (HLM) was used to examine the relationship between PCK Interview
scores and gains in child scores from T1 to T2 (see Table 1). HLM is generally
considered appropriate whenever data has a nested structure, as it does in this study, since
child outcomes are clustered by teacher and teachers are clustered by program sites.
HLM offers several advantages over simple regression analysis in this case. First, while
regression techniques allow examination of relationships between teacher PCK and child
outcomes within classrooms, they neglect to account for variance related to program site.
That is, because we know at which site each teacher is located, if we do not conduct
analyses that account for site-level variance, we are not using all the available
information. At least some of the variance in child outcomes is likely to be due to
program site, rather than teacher; HLM allows us to attribute that variance accurately
while assessing relationships between teacher PCK and child outcomes, yielding greater
precision. Moreover, HLM allows the study to specifically examine the effects of
program site and how those effects interact with teacher effects. That is, we can
PCK for Preschool 22
determine whether and how site characteristics might affect relationships between teacher
language and child outcomes.
In effect, HLM creates one linear regression model at each level of analysis and
then combines these equations into a single model. In this study, child variables (gains in
TEMA-3 scores) are at Level 1, teacher variables (math-related language frequencies and
interview scores) are at Level 2, and while no site variables are used as predictors, the use
of a third level allows an examination of the effects of program sites. In this way, the
modeled “levels” mimic the hierarchical clustering of the actual children, classrooms, and
sites. In Level-1 of the example below, the gains of any particular child (TEMDIF) are
modeled as the mean gains of all children in that classroom (π0) plus some unique effect
associated with the particular child whose gains are being modeled (e). The Level-2
model is meant to predict the mean gains of all the children in a given classroom (π0) as
the sum of the mean gains of all children at that particular program site (β 00 ) plus a
unique effect associated with that particular classroom (r0). Level-3 models the mean
gains of all children at a given program site (β 00 ) as the grand mean of all children’s
gains (γ000 ) plus a unique effect due to program site (u00).
Level-1 Model
TEMDIF = π0 + e
Level-2 Model
π0 = β00 + r0
Level-3 Model
β 00 = γ000 + u00
These 3 levels yield a combined model of:
PCK for Preschool 23
TEMDIF = γ000 + u00+ r0+ e
This explanatory model of gains on the TEMA-3 is sometimes called a fully
unconditional model (see Raudenbush & Bryk, 2002, p. 24) because no predictors (or
conditions) are specified. It is generally helpful to run such a model as a preliminary step
because it provides baseline information about variability in the outcome variable (in this
case, TEMDIF, or children’s TEMA-3 gains) at each level of analysis. Results of the
fully unconditional model for TEMDIF indicate that 19.6% of the variance between child
gains can be attributed to differences at the classroom level. Since this study is focused
on the impact of teachers’ knowledge and practice upon child outcomes, this is the
portion of the variance in child gains that is of interest. The fully unconditional model
also indicates that this variance is significant, X2 = 25.86 with 6 df (p <.000), but variance
between sites is not, X2 = 24.31 with 15 df (p = 0.06, ns). In other words, while
classroom makes a significant contribution to variation among children’s TEMA-3 gains,
program site does not. Teachers (or something at the classroom level) appear to be
having more of an impact upon children’s mathematics learning than program sites.
To evaluate whether teacher PCK for preschool mathematics as measured by the
new teacher interview can explain some of this classroom-level variance, interview
scores were entered in the model as predictors at Level-2 (ITONETWO), resulting in the
following model construction:
Level-1 Model
TEMDIF = π0 + e
Level-2 Model
π0 = β00 + β01(ITONETWO) + r0
PCK for Preschool 24
Level-3 Model
β 00 = γ000 + u00
β 01 = γ010
When interview scores are entered into the model as a predictor, they significantly
and positively predict gains in child outcomes; that is, the higher the teacher’s PCK score,
the greater the gains children in her classroom are likely to have made from T1 to T2 (see
Table 1).
Table 1
Results of 3-Level HLM Models to Predict Child Gains on TEMA-3 (TEMDIF) Preschool Mathematics PCK Interview as a Predictor Fixed Effect Coefficient